The mean is the measure of central tendency most affected by outliers. Because the mean uses every value in a dataset to calculate the average, a single extreme number can shift it dramatically. The median and mode, by contrast, are far more resistant to extreme values.
Why the Mean Is So Sensitive
The mean adds up every value in a dataset and divides by the total count. That formula gives every data point equal weight, including any extreme ones. If you have five tree circumference measurements averaging 176 mm, adding a single outlier of 700 mm jumps the mean to 263 mm, a shift of nearly 50%. The outlier gets folded directly into the sum, dragging the result toward it.
This happens in one direction: outliers pull the mean toward themselves. A handful of unusually large values will inflate the mean. A handful of unusually small values will deflate it. The further the outlier sits from the rest of the data, the larger the distortion.
How the Median Resists Outliers
The median is the middle value when you line up all the numbers in order. It only cares about position, not magnitude. Using the same tree example, adding that 700 mm outlier shifts the median from 177 mm to just 190 mm. That’s a change of about 7%, compared to the mean’s 50% jump from the same data.
The median does still move slightly when outliers are added, because inserting a new value can shift which number sits in the middle. But even if you replace that 700 mm outlier with 7,000 mm or 70,000 mm, the median stays at 190 mm. The size of the extreme value is irrelevant. This is why income statistics, home prices, and other datasets with long tails typically report the median rather than the mean. A few extremely high earners can inflate a mean income figure well beyond what most people actually make, while the median stays anchored to the typical experience.
The Mode Is the Most Resistant
The mode, which is simply the most frequently occurring value, is even more resistant to outliers than the median. Outliers are by definition rare, so they almost never become the most common value. Research in robust statistics has shown that mode estimates are essentially unaffected by outliers, while the median is at least slightly influenced by each one. In practical terms, the mode has a lower “rejection point,” meaning extreme values get ignored entirely rather than nudging the result.
That said, the mode has its own limitations. Many datasets don’t have a clear mode at all, especially with continuous measurements where no value repeats exactly. So while it’s technically the most outlier-proof measure, it’s not always useful.
How Outliers Create Skewed Distributions
Outliers don’t just shift the mean. They also change the shape of the distribution, pulling it into a skewed pattern with a long tail on one side. When this happens, the mean, median, and mode separate from each other in a predictable order.
In a right-skewed distribution (tail stretching toward higher values), the mean sits furthest to the right, followed by the median, with the mode closest to the peak. In a left-skewed distribution, the order reverses: the mean gets pulled furthest toward the low end. The key pattern is that the mean always chases the tail, because it’s the only measure that factors in how far away each value is, not just where it sits in the lineup.
This ordering is a useful diagnostic tool. If you calculate both the mean and median of a dataset and notice they’re far apart, that gap signals the presence of outliers or strong skew. The direction of the gap tells you which way the extreme values pull.
Choosing the Right Measure
Your choice of central tendency depends on whether your data contains outliers and what story you’re trying to tell.
- Mean: Best for symmetric data without extreme values. It uses all available information, which makes it the most precise measure when the data cooperates, but the most misleading when it doesn’t.
- Median: Best when outliers are present or the distribution is skewed. It represents the “typical” value more faithfully in these situations and is the standard choice for things like household income, home sale prices, and hospital wait times.
- Mode: Best for categorical data (like the most popular car color) or when you need the single most common value. It ignores outliers completely but only works well when there’s a clear peak in the data.
If someone reports a mean without mentioning the median, and the data could plausibly contain extreme values, treat the number with caution. The gap between mean and median is often where the real story hides.